The open tarot dataset
All 78 tarot cards of the Rider–Waite–Smith deck, with love and relationship keywords for upright and reversed positions, in 22 languages — plus arcana, suit, element, yes/no polarity and a public-domain 1909 scan per card. One JSON file, no signup, free to use commercially under CC BY 4.0.
What is in the file
One array of 78 card objects. Each carries the stable English slug used as its identifier, the canonical English name, arcana, number, suit, element, a yes/no reading for both orientations, the image URL of the 1909 scan, and a love map keyed by language code.
{
"slug": "the-lovers",
"name": "The Lovers",
"arcana": "major",
"number": 6,
"suit": null,
"element": null,
"yes_no": { "upright": "yes", "reversed": "no" },
"image": "https://www.loveoracle.app/deck/the-lovers.webp",
"url": "https://www.loveoracle.app/cards/the-lovers",
"love": {
"en": { "name": "The Lovers", "upright": "deep union, a real choice of the heart", … },
"ja": { "name": "恋人", "upright": "…", "reversed": "…" },
…
}
}Slugs and English names are identifiers and never change. Elements are given for the four suits only: the elemental attributions of the major arcana differ between schools, and we would rather leave a field null than state a contested one as fact.
Why this file and not another one
Public tarot datasets exist, and every one we could find is English-only. This one carries all 22 languages in the same file: en, es, pt, fr, de, it, nl, da, no, fi, pl, hu, tr, ru, ja, ko, zh, zh-TW, th, vi, hi, id. That is 3432 keyword phrases, which is the part nobody wants to re-do.
Two honest caveats, so you can decide whether it fits. The keywords are written for love and relationship questions specifically, not for general divination — a career-focused app will want different phrasing. And the 21 non-English languages were machine-translated from the English source and have not been reviewed by native speakers; the structure is reliable, the idiom in Thai and Vietnamese may not be.
Licence and attribution
Creative Commons Attribution 4.0. Use it commercially, change it, ship it in a product. The one condition is credit: name Love Oracle and link back.
Tarot data from Love Oracle — https://www.loveoracle.app
CC BY 4.0The card artwork is Pamela Colman Smith's 1909 edition, which is public domain worldwide and carries no condition at all.
Fetch it
curl -s https://www.loveoracle.app/dataset/love-tarot.json | jq '.cards[] | select(.slug=="the-lovers") | .love.es'Served with Access-Control-Allow-Origin: *, so it works from a browser. It is a static file on a CDN — hotlink it if you like, or vendor a copy; both are fine.
Machine-readable access to the rest
Every published page here also answers to Accept: text/markdown, and the same documents sit at /md/… as plain URLs. There is an MCP server at https://www.loveoracle.app/api/mcp exposing three read-only tools over the same corpus — card search, card love meaning, and the long answers to common love questions — so an agent can look a card up without scraping.
curl -H "Accept: text/markdown" https://www.loveoracle.app/cards/the-lovers
curl https://www.loveoracle.app/md/es/cards/the-loversChanges and contact
Version 1.0.0. Corrections to a translation, a card meaning, or the yes/no polarity are genuinely welcome — mail hello@loveoracle.app and a person reads it. If you build something with this, we would like to see it.
The readings themselves are at loveoracle.app, and every card page shows the same data rendered for humans, with a long-form interpretation the dataset does not include.